base patterson, 2004 systematic changes in families following prevention trials

8/13/2019 BASE PATTERSON, 2004 Systematic Changes in Families Following Prevention Trials

1/14

oumal ofAhnomal Child Psychology. Vol 32. No. 6, December 2004. pp. 621-633 (


2/14

62 2 Patterson DeGa rmo and Forgatc

reared by single mothers increased the risk of incarcera-tion as an adult (Mc Loyd. 1998). Furthermore, parental di-vorce had long-term negative implications for adult men-tal health {Chase-Lansdale. \995). Hethcrington (1989)found that in the years immediately following the separa-tion, there was a signilicarU increase in risk fur boys* ag-gres.sion, noncomptiance. and underachieving. These find-ings were replicated in a longitudinal ana lysis by Martinezand Forgatch (2001).

It is important to note, however, that there is goodreason to believe that some of the adjustment difficultiesfor children of divorced families may be linked with par-enting problems that existed ptior to the divorce (Block,Block. & Gjerde, 1986: Shaw. Emery. & Tuer. 1993). Inthis sense, for some fatnilies, the divorce process may am-plify existing dev iancy proce.sses whe reas in other families creates deviancy processe s that did not exist beforehand.Cherlin et al. (1991) examined fmdings from two nationallongitudinal surveys. After controlling for the predivorceadjustment as two-parent families, the data showed thatfrom 11 to 2O'Jf> of the hoys in the now divorced sam-ples were having adjustment problems. Data from passivelongitudinal studies at both Virginia and Oregon showedthat in these divorced families negative child outcomeswere associated with disrupted parenting skills (Forgatch,Patterson, & Ray. 1996; Hetherington, 1989).

As noted earlier, in longitudinal randomized preven-

tion trials with recently separated mothers, ii wa.s hypoth-esized that families in the untreated comparison groupwould show increasing disruptions in measures of par-enting practices and in deviant child outcome. However,families receiving PM T would be expected to show mod-est gains or at least maintain an effective level of parentingpractices, This group wotiid also be expected to show mod-est reductions in levels of child deviancy. For exam ple, inFig. 1. findings from the Oregon Divorce Study (ODS).families in the untreated group showed systematic deterio-ration in parenting practices for both positive parenting (acomposite of monitoring, problem solving, involvement.

and encouragement) and coercive discipline (a compositeof inept discipline, negative reinforcement, and negativereciprocity: see Ma rtinez & Forgatch. 2001 ). Incidentally,the brief increase in coercive discipline confrontations at6 months for families in the E group is a replication ofprior Hndings from a clinical study hy Stoolmiller. Dun-can, Bank, and Patterson (19 93). As the parent confrontsthe child on discipline issues , the parent-child interactionsactually become more aversive for a short time. But the strugg le work through pattern is also a pred ictor for fu-ture .successful adjustment.

The data showed that by Month i 2 . there were al-ready significant differences between the two groups. For

.25-1

.20

.15-

.10

.050

-0 5-.10-.15-.20.-.25-.30

Coercive Discipline

Line Months 12 Months 18 Months 30 Mon

.25-1

.20

.15-

.10-05 -

0-

-.05-- . 1 0 .15--.20--.25--.30

Positive Parenting

6 Months 12 Months 18 Months 30 MonUn e

Experimental

- Control

Fig . 1. The preveniion effecl for effeclive p arenting practices (M Forgatch. 2(K) I).

families in the untreated group , as parenting practicteriorated, child outcomes worsened. In keeping witpectations, a comparison e)f the F and C grotips shthat boys in the E group were significantly better adjon measures of compliance (Martinez & Forgatch. 2school achievement (Forgatch & DeGarmo. 1999)viant peer relations, delinquency, and anti.social beh(Forgatch & D eGarmo, 2002).

The findings are consistent with expectations fovention trials. However, the findings also raise thequestion of why one would expect single-parent famin the comparison group to deteriorate in both pare

practices and child outcomes. As a general case, ies have repeatedly shown that recently separated ers experience dramatic increases in stressors (ForPatterson, & Skinner. 1988; Hetherington. 1989). Ltudinal studies have also shown that increased straccompanied hy disrupted parenting practices (CoPatterson. Ge. 1993: Fo rgatch. Patterson, & Ray. 1Although there are multiple components contributisingle-parent stress, the mo re interesting one is assowith the repartnering p rocess . For example. Forgatch(1996) modeled d ata that showed that frequent repaing was a major contributor to the stress experiencsingle mothers. Frequent repartnering was also asso


3/14

Systematic Changes in Families 623

with growth in child antisocial (Forgatch et al.. 1996). Infact, there was an accumulative effect of prior relalion-ship transitions (DeGai'mo& Forgatch. 1999; Maninez&Forgatch, 2002) on child negative outcomes. In a similarvein, there was a linear relation between transition fre-quency (reparinering) and likelihood of early-onset policearrest (Patterson. 1996). Transition frequency also con-tributes to increasing risk for disrupted parenting (Capaldi& Patterson. 1994; Forgatch et al., 1996).

In Ihe present con text, the hypoihesi.s would be that asubgroup of single mothers engages in frequent repartncr-ing. Presumably, this subgroup contributes signilicantlyto the increasing disruptions in parenting praetiees andconcom itant increases in ehilc deviancy. We suspect theremay be several other trajectories that may lead to a simi-lar outcom e. For exam ple, divorced m others who become

irritable-depres sed are likely to have less

contidant supportover time, which leads to future disruptions in parenting(DeGarm o & Forgatch 1997; Patterson & Forgatch. 1990).Alternatively, one might hypothesize that it is the absenceof a father (igure that leads to children acting out. How-ever, follow-up findings show that children from widowedfamilies do not tend to be aggressive.

Nor is there homogeneity within the interventiongroup of a prevention trial for parents or tor children.Paiiieipating in parent group m eetings suggests smorgas-bord as a metaphor lor tliinking about the postseparationintervention process. The sessions provide an array of op-

portunities where each parent makes her own decisionabout what to work on and when. In some families, prob-lems with the child are just dev eloping. In other families,the parent may be concerned about the child's sadness orachievement or relationship problems with peers: othersmay be concerned with noncompliance or oppositionalbehavior. We hypothesized that at any given point duringPMT intervention, there are only a few parents workingon the same child problems. This is unlike the case forclinical interventions with families of aggressive childrenwhere sam ples are selected because they share a comm onset of symptoms. Within the first week or two, most par-ents in a clinical sample would be focused on some aspectof noncompliance. We believe these differences accountfor the 6-month delay in treatment effects observed in theODS prevention trial (Forgatch & DeGarmo. 2002).

Such diversity of trajectories within a group designprovides an excellent oppo rtunity for evaluating the vari-ance in systems of ehange. Even a casual reading of sys-tems theory (Granic 2000; Lewis & Granic. 2000; ameroff 1989) would suggest that there may be muchmore going on during a prevention trial than is evidentfrom simple pre- and postcomparisons of parenting prac-

tices and child outcomes. According to Sameroff (1989).

one of the prime characteristics of a system lies in its In-terconnectedness. At one level, this may simply be a looseway of saying that family process changes over time in anorderly fashion. It could also be interpreted to mean that.some changes in a family member could correlate withchanges in the behavior of another member during thesame time interval. For example, change in the problemchild might be accompanied by collateral changes in sib-lings as shown by Arnold, Levine, and Patterson (1975).Similarly, changes in child behaviors might accompanyintervention-induced changes in marital relationships.

Our first attempt to explore the interconnectednesshypothesis focused on the changes that accompan ied andfollowed participation in FMT presented during a series ofgroup sessions (DeGarmo. Patterson. & Forgatch, 2004).The analyses were focused on the interplay among mea-

sures of parenting practices, measures of child and mater-nal depression, and a numb er of child behavior problem s.The variables were expressed as growth curves and wereassessed at base line. 6. 12, 18, and 30 m onths. Two typesof collateral change were studied. The tirst was a col-lateral change across members shared across time. Here,chang es in the behavior of one member might ct)vary withchanges in feeling or behavior on the part of the other. Anexample of this kind of collateral ehange was found in apath coeflicicnt of 51 between the changes in boys* anti-social behavior and changes in maternal depression, frombaseline through 30 m onths. The second pattern of collat-

eral changes that was of interest was within-agent c hange sacross time. For example, the individual's feelings mightbe changing during the same time interval during whichthe child's behavior is chang ing. Within individuals, it waspossible that changes in feelings (depression) would fol-low changes in their own behavior. The path coefficientfrom changes in boys' internalizing symptoms to changesin their antisocial behavior was .42. hicidentaily, in ei-ther case if the direction of effects was reversed , the modelno longer fit.

Neither of these results was particularly surprising.What was surprising was that collateral effects such asihese had not been noted before. It makes sense that there isprobably some sort of reciprocal relation between cha ngesin what people do and chan ges in how they feel. DeGa rmoet al. (2004) also presented the data from an E by C com-parison that suggested that the sequence of change itselfmight be orderly. It seemed that group differences in par-enting and child antisocial behavior emerged (irst and wasfollowed by later increasing effect sizes for both the boys'behavior and mothers' depression scores.

The current report examines a further derivative of theorderly sequence conc ept. It is hypothesized that collateral

changes in how the mother feels (depression) may be a


4/14

624 Patterson DeG armo. and Forgatch

PMT InducedChanges Parenting

ChildOutcomes

MaintainEffectiveParenting

Depression

Collateral Changes

1 Within time acrossfamily menibers

2. Within family membersacross time

Sense of

Parent as a

Person

Sense ofSelf-Efficacy

Fig. 2. Theoretical model of collateral changes during PMT.

predictor of later improvements in her parenting practices.

This study extends the D eGan no et al. (2004) analyses intwo different w ays. Their analysis of collateral chang e washased on the eombined information from changes withinthe E group with the changes within the C group. Thecurrent report is focused only on the changes that occurwithin the E gro up . This permits a unilateral focus on PMTinduced change per se. e also assumed that early chan gesin maternal depression might have a special significancefor later stages of intervention (Fig. 2).

Forgatch. Bullock, and Patterson (2004) discussedthe different impact on family life between what parentsdo in terms of parenting practices and how it is that par-

ents think and feel about themselves as a person. Carryingthis distinction into the present context, it is assumed thatPM T alters not only what the parent does in managing herchild but also alters how she thinks and feels about herselfas a parent. Furthermore, it is hypothesized that the collat-eral changes discussed in this study signiticaiitly impactthe various indicators that define the latent construct forParent as a Person. Although this study included measuresfor only the depression indicator, it seems reasonable toencourage other investigators to use a broader specifica-tion. In keeping with a broader perspective. Price's SEM

model (Price, 2003) showed that a sense of mastery

signilicantly related to reduced depression and both munique contributions to long-term re-employment.The indicato r for maternal depression plays a key

in our model for collateral change. Recent findings inresearch literature would lead one to believe that this svariable has also come to play a key role in theories ogression (Reid, Patterson, Snyder. 2002). For exammaternal depression seems to have a bidirectional rtion with parenting practices. Improving parenting ptices produces decreases in depression (Patterson. 1and increasing parental depression produces disruptin parenting (Conger et al., 1995).

Maternal depression also seems to have a biditional relation with child antisocial behavior. As nearlier, separation and divorce were often followed bycreases in maternal depression. These increases, in twere followed by increases in child antisocial behavData from a divorce sample also showed that childviancy contributed to increased maternal depres(Forgatch et al., 1996). As already noted. DeGarmo e(2004) found that negative growth in child antisociahavior over a 30-month interval covaried with negagrowth in maternal depression.


5/14

Systematic Changes in Families 62 5

Even this cursory view suggests that maternal de-pression and child aggression are tied together is someinteresting ways. In the present conlexl. we hypothesizedthai ihi)se mothers who show an early reduction in depres-sion would he more likely to maintain effective pa renting ayear later. Presumably, the earlier this shift in the mother'ssense of being an efficacious parent, the sooner it is thaithe parenting p ractices would improve. In this sense, it isimplied that corollary changes that occur early in preven-tion trials may have some special properties. We examinedthe possibility that early changes in maternal depressionlor families in the E group inay relate to their long-termmaintenance of effective parenting practices. Incidentally,if such a positive feed-forward loop were shown to exist, itwould exemplify what is meant by self-organizing change(Granic. 2000; Sameroff 1984).

There is no strong theory that says that both within-subject, across-time and across-subjeci. across-time col-lateral changes must characterize all PMT prevention tri-als. We view both the hypotheses themselves and the dalain support as exploratory in nature. However, the corol-lary change hypotheses have the virtue of being reasonablederivatives of systems views of family process, and bothviews are eminently testable, given the current data set.

Complete data were available for 153 recently sep-arated families who were randomly assigned to the Egroup. Data were available to examine three sets of vari-ables required to test the hypotheses. composite measure

of parenting effectiveness included observation-based in-dicators. There was also a composite measure of childantisocial behavior that included an observation-based in-dicator as well as a self-report measure of maternal de-pression. The first hypothesis was that during the first12 months, reductions in maternal depression would beshown to be a col lateral change variable that covaried witheither changes in child antisocial behavior or changes inparenting. The second h ypothesis was that the early reduc -tion in matemal depression would identify those motherswho m ade further improvements in their parenting duringthe last 12 months ofthe study. In effect, early changes in

matem al depression function as a feed-forward amplify-ing loop.

M E T H O D S

Participants at baseline were 238 recently separatedsingle mothers and their sons residing in a medium-sizedcity in the Pacific Northwest. Families were randomlyassigned to intervention n = 153) or control conditions(/) = 85). At the 30-months follow-up, the participating

sample was 209 families (88% retention) with no dif-ferences in attrition between the families randomly as-signed to the intervention (/; 133) or control co nditions{n = 76). The baseline families were recruited throughmedia advertisements, flyers distributed throughout thecomm unity, and divorce court records. M others in eligiblefamilies (a) had been separated from their partner withinthe prior 3-24 months, (b) resided with a biological sonin Grades I through 3. and (c) did not cohabit with a newpartner. The sample was restricted to boys, because theyare more likely than girls to exhibit adverse effects ofdivorce as preadolescents (Hetherington & Clingempeel.1992; Shaw. Emery. & Tuer. 1993).

At baseline, mothers had been separated for an av-erage of 9.2 months. Families tended to be small, with2.1 children on average. The mean age ofthe mothers was34.8 years (5D = 5.4; range 21.4-4 9.6), and the mean agefor the boys was 7.8 years {SD = .93; range 6.1-10.4).The racial or ethnic composition of the boys in the sam-ple was 86% White, 1% African American, 2% Latino.2% Native American, and 99f from a mixture of racialor ethnic groups. This distribution reflected the racial orethnic makeup of the community in which the study wasconducted. The mean annual family income was S14.900.which was similar to ihat reported for other female-headedhouseholds with children in the county at that time (i.e., 15,300: US Department of Comm erce. Bureau of th eCensus. 1993). Seventy-six percent of the families werereceiving public assistance.

Most mothers (769f) had some academic or voca-tional training beyond high school, although only 17%had completed a 4-year college degree or higher, Ap-proximately 20% of the women completed their educa-tion with high school graduation, and 4% had not com-pleted high school. Most mothers were classified withinlower- and working-class ranges in terms of occupation(Hollingshead. 1975): 32% unskilled, 21% semiskilled.23% clerical or skilled, 22% minor professional to mediumbusiness, and 3% major business or major professional.

At baseline, the groups differed in terms of months

since separation, r(236) = 3.21./) .001. and the boy'sage. r(236) = 2.27, p = .02. On average, mothers in theE group M = 9.84), SD = 5.79) had been .sepatated forabout 2.4 months longer than those in the control group( M = 7 . 4 8 . SD ^4.56). Boys in E group (M = 7.65..SD= .93) were about .28 years younge r than those in the con-trol group M = 7.93, SD = .92) on average. Models weretested controlling for these variables. None of these con-trol variables altered the substantive findings of the presentanalysis, thus they were excluded from the presentation ofresults.


6/14

626 Patterson. DeGarmo and Forgatch

Study Design

The study employed an E longitudinal design. Fam-ilies were assigned randomly with approximately two-thirds to the E group (/( = 153) and one-third to theno intervention C group {n = 85). The unequal assign-ment to group condition was done to provide sufficientsample size within the E group to examine po tential full-implementation effects of the intervention (Vinokur. Price,& Kapliin. 1991). Mothers in the E group were invited toparticipate in the intervention; and families in the C con-dition received no intervention. The intervention involvedno work with the children. Only mothers participated.

Families in both groups received extensive m ultiple-method, -setting, and -agent assessment live times on thesame timeline: baseline, 6. 12. 18, and 30 months. A minor

assessment, including telephone interviews with mothersand ratings from teachers, occurred at 24 months but didnot include assessment of the key outcomes analyzed torthe present paper. Currently, assessment continues and isnow being collected at 48 months. All E families hadcompleted the intervention by the 6-month assessment,although 4-6 weeks may have transpired between termi-nation ot the intervention and the 6-month assessment.

Intervention

The intervention consistedof series of parent group

meetings held weekly in the early evening hours at theOregon Social Learning Center (OSLC). The original in-tervention curriculum included 16 weekly topics, but twotopics were combined with others for parsimony when theintervention was underway. Thirteen parent groups rangedin size from 6-16 (M 9.5). E-group m others participatedin an average of 8.5 sessions {SD = 5.7).

The content of intervention sessions, description ofinterventionist training, and details regarding program fi-delity are provided elsewhere (Forgatch & DeGarmo.1999). The m anualized intervention program is describedin Pareniing Throu^ih Change (Forgatch, 1994). The pro-

gram also included a 30-min videotape.771^* Divorce Worhmi (Forgatch & Marque z. 1993). which shows threefamilies using effective parenting practices to help theirchildren adjust to the divorce transition.

The intervention was built around five theoreticallybased effective parenting practices (i.e.. appropriate dis-cipline, skill encouragement, monitoring, problem solv-ing, and positive involvement) and olher issues relevantto divorcing women (e.g.. regulating negative emotionsand managing interpersonal conflict). The parenting top-ics taught mothers strategies for decreasing coercive ex-changes with theirchildren by responding early and appro-

priately to child misbehavior with noncorporal disc(e.g., time out, work chores, and privilege removamultaneously, the intervention focused on the use otingent positive reinforcement (e.g., praise and inccharts) to promote prosocial behavior. The topics w esented in an integrated step-by-step approach. Eactopic was introduced to build on previously learnedTopics were usually introduced in one or more seand then reviewed and revisited throughout the remof the program. The program was flexible in that il alparticipants to discuss current relevant issues as partregular agenda for each session. These issues werelinked directly to a specific curriculum topic.

Measures and Construct Scores

Multiple-informant and multiple-method aments were conducted at each assessment. Relevantions of the assessment battery for this study incstructured interviews wilh mothers and laboratory vations of mother-child interactions during structurteraction tasks. Participants were paid approximateper hour for their time. Laboratory observation inclset of structured interaction task s totaling 45 m in. Ttivities included: four 5-min mother-son problem-sdiscussions about current hot conflicts (e.g.. chores, problem s, behavior), a 10-min teaching task (involvmother assisting her son with a series of academiclems set one grade level above his current gra de), a unstructured activity {involving directions for the cnot play with certain toys and to put toys away), anmin period for participants to share refreshments. interactions were videotaped and scored using thepersonal Process Code (lPC; Rusby, Estes. & Di1991) and a global rating system (Forgatch. KnutsMayne. 1992). Approximately \5 of the interactiwere scored by randomly selected pairs of coders sess intercoder agreement. Average Cohen's KappIPC content codes was .78 (range = .11 to .80) and affect was .70 (range = .67 to .76) across all five w

To model change over time, mean construct were created for outcomes comprising multiple indusing methods outlined by Stoolmiller (1995) for ing growth of construct scores. The focus in this reon antisocial behaviors of the boy, maternal depreand effective parenting practices.

Boy Antisocial Behaviors

Child antisocial behavior was a composite conscore of three m ain indicators, consisting of ob serv


7/14

Systematic Changes in Fam ilies 6 7

measures of noncompliance and aggression and a mother-reported standardized measure of delinquency. The three-indicator construct produced Cronbach's alphas of .59..68. .6 3, .64, and .6 5. respectively.

Noncompliance was a construct score itself devel-oped and validated in Martinez and Forgatch (2001). Thenoncom pliance score included three indicators: a microso-cial score, a coder rating scale score, and an interviewerimpression score. The indicators were subjected to prin-cipal components factor analyses yielding a single factorsolution at each wave. The factor loading ranged from.65 to .89 across the five waves of data. The micro socialscore was computed from IPC codes. The score was theproportion of child noncompliance (relative to all olherbehaviors) following a mother's directive, that is. the totalnumber of child non compliance divided by the total num -ber of directives issued by the m other during the o bserva-tion. This proportion score accounts for differential baserates of matenial d irectives, which could arbitrarily inflateor deflate child noncompliance rates. The intraclass corre-lation coefficient (ICC) of intercoder reliability for the mi-crosocial noncompliance score was .71. .80. .84, .64. and.86. respectively. Coder ratings was a scale score basedon global observations m ade directly following scoring ofthe microsocial behaviors. Ten Likert-scale items assessednoncompliance on a scale from to 5. Sample items in-clude: withdrew from interaction in a negative way. wasnoncompliant, complied with commands from mom. wasresistant to doing task. Cronbach's alphas were .90, .89,.87. .90. and .89: ICCs were .7 7. .87, .87 . .66. and .90, re-spectively. FinaUy,ihechiUl interviewer impre.s.sions ficorcwas based on global ratings of noncompliance. This scaleconsisted of three items rated from 1-5 including itemssuch as: was cooperative during the interview, was non-compliant with you. and was noncompliant with mother.Cronbach's alphas were .85 . .84. .85. .81. and .83. respec-tively.

Aggression w as a microsocial indicator scored as therate per minute of physical aversive behaviors observedduring the structured interaction tasks. Examples of phys-

ically aversive behaviors from the IPC included any con-tact of the child with the mother involving hitting, kicking.pinching, and so on. The aggression indicator was log-transformed before factor analyses, because the indicatorwas skewed and leptokurtic in its distribution.

Delinquency was measured with the delinquency Tscore from the mothers' report on the Child BehaviorChecklist (CBCL: Achenbach. 1992). The 7"score is na-tionally normed and consists of 13 items describing childbehaviors over the previous 2 months. Items were ratedon a 3-point scale from 0 not true) to 2 veiy true or of-ten true). Sample items include the following: steals, uses

drugs or alcohol, or vandalism. Cronbach's alphas were.64, .63 , .72, .73 . and .75 across waves.

Maternai Depression

A single indicator defined maternal depression. Ma-ternal report of depressed mood was assessed by the scalefrom the Center for Epidemiological Studies of Depres-sion CESD:Radloff. 1977). This was a 20-item symptom-oriented report, where each item was measured on a4-point scale (0-3) indicating the frequency of events dur-ing the previous week. The events ranged from rarely ornone (0-1 day) to masv or all of the time (5-7 days).Sample items included felt depressed, fearful, lonely, orhopeful about the future. Cronbach's alpha reliabilitiesranged from .92 to .94 as assessed from baseline through30 months.

Effective Parenting

Domains of effective and ineffective parenting werederived from theoretically based research on Patterson'scoercion model (Patterson. 1982; Patterson. Reid, &Dishion, 1992). These observational I y based measuresof effective skills and ineffective coercive strategies havedemonstrated sensitivity to change in PM T and have shown

convergent and predictive validity in at-risk and divorcesamples (Eorgateh et a ., 1996: Forgatch & DeG armo .1999; Patterson. 1982: Patterson et al., 1992). Eor thisanalysis, the final construct score was the mean of theseven indicators of parenting skill, four positive measures,and three coercive measures. Positive domains includedpositive involvement, skill encouragement, problem solv-ing, and m onitoring. Negative dom ains included inept dis-cipline, negative reinforcement, and negative reciprocity.Negative domains were reverse scored, and each indica-tor was rescaled to range from O-I so that a higher scorereflected effective parenting. The seven-indicator growth

construct produced Cronbach's alphas of .73, .70, .73. .72,and .63, respectively.Positive involvement was a mean scale score of coder

ratings following each of the structured interaction tasks.Eive indicators of mother's observed prosocial involve-ment following the interaction tasks and two overall rat-ings were averaged for the final scale. The seven indicatorsincluded 38 Likert-scale items concerning the mother's in-teraction with her son (e.g.. showed empathy, supp ort, gen-uine concern; provided encouragement; showed warmth:showed respect; was accepting: and w as affectionate). Al-phas for the overall scale ranged from .90-.94 across the


8/14

628 Patterson DeGarmo and Forgatc

assessment waves. ICCs were .83. .90. .82, ,79. and .93from baseline to 30 months.

Skill encottrctf^emeiu was a mean scale score hased onglobal coder ratings following the 10-min leachinii task.Eleven .scale items a.ssessed the mothers" ability to pro-mote child skill development through contingent encour-agement and scatTolding strategics (e.g., breaks task intomanageable steps, reinforces success, prompts appropriatebehavior, and corrects in a nonaversive way). Cronbach"salphas were .69. .73. .81. .70 and .67: ICCs were .7 3. .67..66, .48, and .76, respectively.

Problem solvitig wa.s a scale score of global coder rat-ings made iolluwingeach ofthe three problem-solving in-teractions involving a mother-identified i.ssue. Nine Likert-scale items were averaged to comp ute the scale score (e.g..solution quality, extent of resolution, likelihood of followthrough, apparent satisfaction). Cron hach's alphas rangedfrom .87 to .92 across the three topics and five waves;ICCs were .77. .81 , .76. .84. and .79. respectively.

Monitoring vwas a scale score from two agents' re-ports of how effective the mother was in supervising herson. Parent interviewers provided global Likert-scale rat-ings on three items: skillful in supervising during the as-sessment, keeps close hack ofyo ungster outside of labora-tory, and skillful at obtaining infoiination from lhe child.Cronbach's alphas were .82, .76. .73. .74, and .60 frombaseline to 30 mon ths. Coders also provided global ratingsof monitoring alter coding the entire 45 minutes oi lab-oratory interactions. The two Likert-sca e items assessedthe mothers' apparent knowledge of their sons' day-to-day activities and tolerance of negative behavior. The twoitems were correlated .22 , . i 3, .38 , .59, and .47 from base-line to 30 months and were significant at the p < .05 ateach wave. The final score was an average of the ratingsby the two agents.

Inept discipline was a scale score based on coderglobal impressions of discipline following their codingthe entire 45 minutes of laboratory observation. The scaleconsisted of 13 items rated from I {itutrue/doesn i fit) to5 lifue/perfeclftl). Samp le items include overly strict, au-

thoritarian, oppressive; erratic, inconsistent, haphazard:threatens unlikely disciplines: and uses nagging or natter-ing to get comp liance. Cronbacli's alpha ranged from .91to .92; ICCs were .70 , .85 . .78. .77. and .88 , respectively.

Nei^ative reinforcemeni was a microsocial scorebased on IPC coding ofthe mother-child interactions ob-served from the full 45 min of interaction. Our operational-ization of negative reinlbreement was based on conflictbouts, which involved an exchange of aversive behaviorbetween mother and son. Negative reinforcement was de-fined as the frequency of conflict bouts initiated by themother and terminated by the son. In the first step of the

bout, the mother introduced an aversive behavior folling a period of at least 12 s of interaction without averbehavior. In the second step, the son responded withaversive behav ior within 12 s. In the last step, the bended following a period of at least 12 s without eiperson's use of aversive behavior. In this sense, the escaped an aversive situation by employing an averbehavior and was negatively reinforced. The ICCs w.78. .58. ,49. .68. and .60. respectively.

Ne >uiive reciprocity was a microsocial score bon the entire 5 minutes of laboratory observation. The rived score was the Haberman binomial : score (Gottm& Roy. 1990) retleding the conditional probability the mother reciprocated the son's aversive behavior wan aversive behavior of her own. ICCs were .65. .74,.63, and .67 from baseline to 30 months.

RESULTS

Analytic Strategy

The analysis plan was to begin by examining invention effects on the separate outcomes first and texamine an interconnected process model of systemchange over time. For the final model, within the famparticipating in P MT g roup sessions, we specified a crlagged panel model to evaluate the interdependencechange among the variables (see Curran & BoIIen, 20For the longitudinal participants, we estimated the muvariate models using full information maximum tikelifestimation (FIML). Assuming that data are missing cpletely at random. FIML is more statistically efficthan listwise deletion methods for longitudinal data Wothke, 2(X)0). Littl e's test for missing d ata indicated FIML was appropriate as the data were missing cpletely at random (x" - 78.847, d f= 76 , p ^ .38).

Given the central role accorded the matemal depsion variable, we began by exam ining chang es in this vable during lhe course of the study. As presented in Fig

the data showed that the situation for the C group falies is unstable; whereas mothers in the treated groupshowing significant reductions in maternal depression.evaluated change in the variables of focus using repemeasures analysis of variance (MANOVA) using polymial contrasts adjusting for the uneven time spacingassessments. In these particular MANOVAs, / tests copare fitted slope coefficients between groups for eachthe transformed time effects. For maternal depressionLinear group x Time effect was significant using intenttreat directional hypotheses for the expected interveneffects (/= 1.89 /; = .06) .


9/14

Systematic Changes in Families 6 9

IB -

17

16

15

14

o 13

12

. Control

I Experimental

Base 6Line Months

12 18 3Months Months Months

Time

Fig. 3, Changes in maierniil depression.

The next step in the analyses was to examine changesin the composite antisocial behavior score. As noted ear-lier. DoGarnio et al. (2004) based their analyses on thecombined E and C group data. They found that over a 30-month interval that reductions in child antisocial covariedwith ch anges in maternal depression.

The findings describing changes in hoy anti.socialbehavior are pre.sented in Fig. 4. On the average across30 months, there was both a reduction in child antisocialbehavior within the E group and at the same time an in-crease in antisocial beha\ ior for the control group. TheLinear group x Time effect was significant / = 2.38, p

.01). It can be seen that during the first 12 months ofthe study, there was an appreciable reduction in boy anti-social tor members in the E group. During that same lime

Experimental

BaseLine

6Months

12Months

Time

18Months

30Months

Fig. 4. Changes in boy antisocial behavior.

interval, the boys in the C group seemed to worsen. How-ever, the hypothesized covariation between changes in boyantisocial with changes in maternal depression during thesame 12-monlh interval was found to be nonsignificant( - . 03 ) .

As noted eariier. the research literature offered strongsupport for a bidirectional relationship between mater-nal depression and parenting practices. This implies thatchanges in either variable could produce changes in theother. In the present context, hypothesizing a feed-forwardloop would require that the direction of effects move fromchanges in parenting to changes in matenial depression.As shown in Fig. I. there were improvements in coercivediscipline p ractices during the first 12 months. It was hy-pothesized that during this interval, changes in parentingpractices would covary with changes in maternal dep res-sion, and the changes in parenting practices would be as-sociated with changes in child antisocial behaviors.

To understand how induced changes in the familysystem worked for the intervention group, we examinedthe interconnecledness of mother s depression, her parent-ing behaviors, and behaviors of the child by cond ucting across-lagged panel model examining changes from base-line to the first year and then unique ch anges from thefirst year to 30 months within the E group. Figure 5 sum-marizes the findings in the fomi of a structural equationspath model using standardized beta coefficients that exam -ine these hypotheses. Although cross-lagged paths weretested among the three main variables, only significantpaths were displayed in the final trimmed model. N o sub-stantive difierence was found between a trimmed modeland one with full cross-lagged paths.

The final process model provided the best fit to thedata, X- 14) - 12.06./? - 60 . CFI 1,00, Before ex-amining the collateral change hypotheses, it is importantto note that the model also includes data that relate to theover-all purpose of the prevention trial. There is a highlysignificani path (B = .37, p < .001) between changesin parenting at 12 months and changes in boys antisocialduring that same time interval. In the same vein, the pathfor changes in p arenting at 30 months to changes in boysantisocial behavior was -.26 { < .01). Clear evidence isprovided that the interventions were doing what they weredesigned to do.

In the general model, it was also the case that thebaseline measu res for parenting effectiveness correlated inthe expected fashion with the baseline measure of depres-sion r 23, p < .01). What is important in testing thefirst hypothe,sis is that changes in parenting during tho.sefirst 12 months covaried significantly with reductions inmatemal depression r 19. p < .05), The findingsshow that within families in the group, there is low-level


10/14

63 0 Patterson, DeG armo , and Forgatch

Bas e Line 12 M onths 3 M onths

A/=153cmin = 12.06, df 14, p = .60, cmindf = .86, cfi = 1.00 p < . 0 5 , *p< .001

Fig. 5. Process model of PMT-induced ch anges williin Ihe ex peri menial group.

covariation between these two variables even during thefirst 12 montbs. Improvements in parenting were associ-ated with collateral reductions in maternal depression formothers in tbe E group. The path coefficient does not.of course, prove that tbe direction of effects moves fromchanges in parenting to cbanges in depression or vice versafor correlational effects.

A separate model was u.sed to test for the significanceof either bidirectional path between maternal depressionand ber parenting. Tbis was done by specifying recipro-cal effects and removing the prior cross lags. Tbe analysessbowed a nonsigniticant path from depression to parenting 6 = .03). The re was a borderline path from chang es inparenting to cbanges in depression R .11). Althoughthe trends are in keeping with the hypothesis, the fact thatthe path from parenting to depression was only of bor-derline significance weakened any argument that requireda particular direction of effects. For the moment, we canonly conclude that there are associations between mater-nal depression and parenting at baseline and 12 months.However, further tests of process hypotheses using laggedsequence can inform the expectations of how changes intbe family system can be associated with future changesin the same system.

Our feed-forward hypothesis requires only that schanges occurring during the lirst 2 months predict whfamilies continued to improve their parenting duringlast 18 months of the study. We hypothesized that aduction in maternal depression during the first 12 mowould serve this function. To provide the most convative estimate possible, the contribution of matemalpression to future parenting m ust remain significant partialling out the prior contribution of parenting. Inmodel, the path coefficient from baseline measures of enting to changes in parenting at 30 months was .35. path from changes in parenting at 12 months to chanin parenting at 30 months was .37. Both of these conbutions were partialled out to arrive at a residual estimof the contribution of earlier changes in maternal depsion. The residual contribution of the prior reductionmatemal depression was a small but statistically signcant one as indicated by the path coeOicient of .14 p05). The findings are consistent with the idea of an earchange in depression contributing to future improvemin parenting. The pattern of findings is also consiswith the status of the matemal d epression variable servas a feed-forward loop providin g uniqu e contrib utionchanges in parenting that occurred postintervention.


11/14


independent contribution of reductions in maternal de-pres.sion were prediclive of unique changes in parentingfrom 2 to 30 months Ihat. in turn, predicted unique chan gein child antisocial behavior from 12 to 30 months.

DISCUSSION

The findings make a case for the general assumptionthat prevention and clinical trials may hoth be character-ized by orderly changes among family members that hasbeen largely overlooked because of our historic preoccu-pation with changes in symptoms. It wa.s also the casethat in many clinical studies, assessments were limited tojust two time points, pre- and postintervention^hardlya sufticient basis tor estimating collateral changes. How-ever, given repeated assessments, both the publication byDeGarmo et al. (2004) and the current findings suggest thatPM T ean set complex processes in motion. Change.s in oneperson ean alter the feelings and behaviors of other familymembers. The data also support the idea that there maybe collateral changes wiihin individuals across time (i.e.,changes in individual behaviors can eovary withchanges in how the individual feels).

The current data suggest that during the tirst12 months, one change that occurs that is fundamentallyimportant is the reduction in maternal depression. Thefindings are consistent wilh the idea that some early im-provements in parenting practices are responsible for theimprovements in maternal depression. However, the find-ings are such that we cannot make a compelling argumentthat the direction of effects is from parenting changes tochanges in depression. For example, we cannot stronglyreject the argument that it was improvements in mater-nal depression that led to the improvements in parent-ing. Perhaps for many, simply being enrolled in a struc-tured program is associated with a reduction in depression(i.e.. now there is renewed hope).

Whether or not the current findings provide an ac-curate specitication of the model for collateral changesin depression, the findings say that given sucb changesoccur early in the treatment process, they served as afeed-forward loop. The chang es in depression that occurearly on in intervention are significant p redictors for futurechanges in parenting. Presumably, if measures of mastery,parental efficacy, and depression were all available, theresulting latent construct would provide a more sensitiveregister of collateral effects. The improved specificationof the model would be further enhanced if the study alsoincludes repeated assessment of all family members.

We need to develop a theory about intervention pro-cess that details the sequence of variables that change

during prevention trials. Who chang es first? Why do theychange? For exam ple, observation data could be collectedon a monthly basis during and following intervention.Given thaf each family memb er was targeted in each obser-vation session, it would be possible to trace the sequenceof change. For example, does il move from changes in theproblem child to changes in siblings and parents.'

Early studies of resistance in clinical samples werelimited examples of this kind of process study. Datashowed that parent depression, social disadvantage, andantisocial behavior were all associated with increased re-sistance to change in parenting and less improvement inchild outcome (Patterson Chamb erlain, i98 8, 1994;Patterson Forgatch. 1985). Micro social analyse sshowed that as the parent attempted to sel limits on theproblem child, child coercive behavior actually increasedon the short run. This effect was also obtained in the For-gatch studies (see Fig. 1. Discipline). Stoolmiller et al.{1993) suggested that successful families followed a dif-ferent resistance trajectory than did nonsuccessful ones.

Our discussions about therapy process have oftencentered on the idea that proces.ses could be generatedby the interventions that serve as feedback loops. Pre-sumably, the function served by these loops could be toenhance (or retard) the process of change. Perhaps theseimaginary loops would also determine long-term mainte-nance. The problem was that the kind of data required wasnot available in most clinical stud ies. Ther e was simply noway to study what takes place among family members astreatment and follow-up phases unfold. The ODS designwith its repeated measures partially .solved that problem.However, several limitations in this study should be noted.One of the biggest handicaps in specifying the model isto be found in the use of a single indicator, maternal de-pression, to define what a mother thinks and feels aboutherself as a parent. In current work with an independentdivorced sample where multiple indicators were available(see Forgatch et al., 1996), we have since found that in-deed maternal depression loaded (.77) on the construct;but so did self-report measures of mastery (.76) and selfesteem/self-worth (.87). A second limitation lies in therestricted number of time points available for a defini-tion of growth or change. We had only three points toassess early and three for late change. Tbe follow-up ofthe ODS study continues; and data are now available at48 months.

Il is obvio us that we are just b eginning to u nderstandhow PMT functions as an agenl of change in the familysystem over time. Following these leads will require majorchang es in the timing and in the range of variables assessedboth during and following prevention trials. New studieswill have to incorporate many features ol the Forgatch


12/14

632 Patterson, DeGarm o, and Forgatch

studies with a focus on data collected at 6-month intervalsand a general sampling of multimethod, multiagent deti-nilions for key constru cts. As noted earlier, we expect thatluture studies will find that shifts in measures of how thefamily members feel and how tbey think about themselves,and other members will play significant roles in these yetto be discovered feed-forward loops.

ACKNOW LEDtiEMENTS

The authors acknowledge suppo rt provided by grantsMH 38318 and MH 5470 3 from ihe Child and A dolescentTreatment and Preventive Intervention Research Branch,DSIR. NIMH, U.S. PHS; Grant No. DA 16097 from thePrevention Research Branch. NIDA. U.S. PHS; and GrantNo. P50 MH 46690 from the Prevention and BehavioralMedicine R esearch B ranch. Division of Fpidemiology andServices Research Branch, NIMH, U.S. PHS.

REFERENCES

Acfienbucli. T. M. (1992 ). Revised CliiUlBehavior CfwcUisl Burlington:University ot Vermoni.

Arnold. J. E.. Levine. A. G. . Patterson. G. R. (1975). Changes in sibtingbehiivior following family inlervention. Journal ofConsulling amiCliimul P.s\rhoiof;y. 43. 683-688.

Bien, N. Z.. & Bry. N. H. (1980). An experimenlally designed com-parison of lour intensities of schoi>l-based prevention programs foradoiescenls wilh adjustment prob lems. Joiinial of Comm unity Psy-clwlo.iiy.^. l l ( ) - l 6 .

Block. J. H.. Block. J.. Gjerde. P. F. {1986). Tbe personality of childrenprior lo divorce: A prospective sludy. Child Developuwiil. 57 827-840.

Capaldi. D . M.. & Patterson. G. R. (1994). Interrelated influences of con-texlual fat-tors on antisocial behavior in childhood and adolescencefor males. In D. Fowles. P. Sutker, S. Goodman lEds.), Pmgrcss inexpeniiienuil personality and psychopmhoiogy research (pp. 16 5-1981. New York: Springer.

Chase-Lansdale. R L. (1995). The long-term effects of parental divorceon the mental health of young adults: Adeveloptne ntal perspeclive.Child D evelf'pmeni. rtfi' 1614-1634.

Cherlin. A. J.. Furstenberg, F. F.. Jr., Chase-Lansdale, P. L.. Kiernan.K. E., Robins, P. K., Morrison. D. R.. et al. (1991). Longitudinalstudies of elTects of divorce on children in Great Britain and theUniled States. Science. 252. 13S6-1389.

Conger. R. D .. Patterson. G. R.. G^. X. (1995). It lakes two to replicate:A mediational model forihe impact of parents' stress on adolescentadjustment. Child Dcvt'lapmenl. 66. 80-97.

CuiTdn.P.J..& Bollen.K, A, (2(K)n.The beslofboth worlds: Combiningautoregressive and latenl curve models. In L. M. Collins & A. G.Saver (Bd^.). New methods for the tiiuilyxis ofdianê: Decade ofbehavior {'(>\>. 107-135). Washington DC: American PsychologicalAssociation.

DeGarm o. D. S.. & Forgaich, M. S. (1997). Conlidani support and mater-nal distress: Predictors of parenting practices for divorced mothers.Personal Relationships. 4. 305-317.

DeGarmo. D. S., & Forgatch. M. S. (1999). Contexts as predictorsof changing malcnial parenting practices in diverse family struc-tures; A ôcial interactional perspective to risk and resilience. InE. M. Hetherington (Bd.). Coping with divorce, single parenting

and remarriage: A risk and resiliency perspective (pp. 227-252Hillsdale. NJ:Erlhaum .

DeGarmo. D. S.. Palterson. G. R.. & Forgatch. M. S. (2004). Whyparent training inter\'enlion outcom es maintain or wane over tiPrevention Science. 5. 73-89.

Eisensladt. T. H.. Eyberg, S. . McNeil, C. B., Newcomb. K..

FLmderburk. B. (1993). Parenl-child interaction therapy withhavior problem children: Relative effectiveness of two stages overall ireatmcnl outcome, .loiinuil of Clinical Chiki Psycholo2 2 .4 2 -5 1 .

Forehand. R., & L ong. N. (1988). Outpatient treatment of the actinchild: Procedures, long-term fnllow-tipdata. and clinical problAdvances in Behaviour Research and Therapy. JO. 129-177.

Forgatch. M. S. (1994). Parenting through change: A piiigraninieiltervention curriculum fur groups of single m others. L'npublishmantiscript, Oregon Social Learning Center. Eugene.

Forgaich. M. S.. BulliKk. B. M.. & Patterson. G. R. (2004). Fromory to practice: Increasing effective parenting through riile-pThe Oregon model ol" parent management training (PMTO l. InSie'mer.(Eds.). Handbook of mental he dth iiitenentions in ihihirand adolescents: An integrated developmental approach, (pp. 78812). San Francisco: Jossey-Bass.

Forgatch. M. S., DeGarmo. D. S. (I999K Parenting through chanAn effective preveniion program for single mothers. Journal Consulting and Clinical Psychoh?gy, 67. 711-724.

Forgatch, M. S.. DeGarm o, D. S. (20()2(. Extending and lesting thecial interaction learning model with divorce samples. In J. B. RG. R. Patterson. & J. SnydcT{As.). Aniisocialheluivior in childrand adolescents: A developmental analysis and model for ivetiiion (pp. 235-256). Washington. DC: American PsychologAssociation.

Forgateh. M. S.. Knutson. N.. & Mayne, T. (1992). Coder impivs.iioof ODS lab tasks. Eugene: Oregon Social Learning Center.

Forgaich, M.S.. Ma rque /, B.( 1993). r/wrf/rwavH'orioMf I Video tEugene, OR: Oregon Social Learning Center.

Forgatch, M, S.. Patterson. G- R.. & Ray. J. A. (19961. Divorce aboys' adjustment problems: Two paths with a single model. IM. Hetherington & E. A. Blechman (Eds.), Stress, coping, aresiliency in children and families (pp. 67-105). Mahwah, Erlbaum.

Forgatch, M. S., Patterson.Ci. R.. Skinner. M. L, (1988). A mediatiomodel for the effect of divorce on antisocial behavior in boyE. M. H etherington J. D. Arasteh (Ed.s.), Impact of divorce, sinparenting, and step-parenting on children pp. 1 35 1 54 . Hi tlsdaNJ: Erlbaum.

Gotiman, J. M.. Roy. A. K. (1990 ). Sequential analyses: A guidebehavioral researchers. New York: Cambridge University P res

Granic. I. (2000). The self-organization of pareni-child relations: Bebidirectional models. In M. D. Lewis & 1. Granic (Eds.l. Emotidevelopment and self-organization: Dynamic .sy.stems approacto emotional development (pp. 267-297). New York: CambriUniversity Press.

Hetheringlon. E. M. 0989). Coping with family transitions: Win

losers, and survivors. Child Development. 60. 1-14.Hetherington, E. M., & Clingempeel. W. G.I 1992). Coping with mtransitions: A family systems perspective. Monographs of the Sciety for Re.\earch in Child têvelopment. 57(2/3. Serial No. 21 242.

Hollingshead. A. B. (1975). Four factor index of social status. Unplished manuscript. Departmen t of Sociology, Yale University, Haven. CT.

Kazdin. A. E. (1997). Parent management training: E vidence.o utcoand issues. Journal of American Academ v of Child (md AdolescPsychiatry. 36. 1349-1356.

Lewis. M. D., Granic. I. (Eds.) (2(XM)l. Emotion, development .self-organization: D ynamic .systems approaches to emotionalvelopment. New York: Cambridge University P ress.

Manin ez. C. R.. .Ir.. & Forgatch. M. S. (2001). Preventing problemsboys" noncompliance: Effects of a piu-ent training interventio


13/14


cts. Journal of Com uliing ami CiiniuilP.'i-y. 416-428.cz . C. R.. Jr. Forgaich, M. S. (2002). Adjusling lochan ge: Link-

ing fiimily siructure transitions with parenting and child adjustmeni.Joiiiiuilof raiiiily P.miiolofiy. 16(2). 107-1 17.

McLoyd. V. C. (I 99 8|. Children in poverly: Developmeni. public policy,

and praciice. In W. Damon. I. E. Sigt-I & K. A. Renningcr (Eds.),Handbook of child p.syrhnhgy (5ih ed., Vnl. 4). New York: Wiley,Patterson. G. R. (1980). Mothers: The unaeknowledged victims.

Moniigtaphs of ihe Socielxfor Reseairh in Child Developmeni,45(.S. Serial No. 186).

Patterson. G. R. (1982). A social leiiniiijf; appmuch: Vol. 3: Coercivefiiiiiily pivcess. Eugene. OR: Casiiilia.

Patterson. G. R. (199ft). Some characteristics of a developmental theorylor early onset delinquency. In M. F. Lcn/enweger. J. J. Haugaard(Eds.). Fyoniiers of devehpmenial psychopalhohgy (pp. 81-124).New York: Oxford University Press.

Patierson. G. R.. & Chamb erlain. P. (I9 88 |. Treatment process: A prob-lem at three levels. In L. C. Wyitne [Ei\.). Siaie of ihe an in familyiherapy reseuivh: Coiumversie.\ and recommendations (pp. 189-223). New York: Family Process Press.

Patterson. G. R.. & Chamb erlain. P. (1994 ). A functional analysis of

resistance during parent training therapy. Clinical Psyrholofty: Sci-ence and Practice. I (11. 3.U7O.Patteisun. G. R.. & Reischman, M. J. (1979). Maintenance of treat-

ment eflects: Some considerations concerning family systems andk-iWm-ayi dM-j,. H ehuvior Therapy. 10. 168-IH5.

Palterson. G. R., & Forgatch. M. S. (1985). Therapist behavior asa detenninant for client resistance: A paradox for the behaviormodified. Jo iirntil of C and Clinical P.sychology, 5. 237-262.

Patterson. G. R.. & Horgatch. M- S- (1990). Initiation and maintenance ofprocess disrupting single-mother families. In G. R. Patterson (Ed.),Depression and aggre.Ksion in family interaction (pp. 209-245)Hillsdale. NJ: Erlbaum.

Patterson. G. R.. Reid, J. B.. & Dishion. T. J. (1992). A .wciul inlerac-tional approach: Vol. 4: Anlisocial hoys. Eugene. OR: Castatia.

Vnc';.R.\\.[.2{){)}'J\inc). Inien'enlion OS hypothesis: Leurninfi from fieldc.xperintenial in ihe Winning New Job.s' pn>f>iant. Paper presentedat Family Conso rtium III. ' Prevention and Dissemina tion. SantaAnna Pueblo. NM.

Radloff. L. S. (1977), The CES-D Scale: A self-repori depression scalefor rese:uch in the general population. Applied P.s\chological Mea-surement. I. ?

Reid, J. B.. Patterson. G. R.. & Snyder. J. (2002). Anlisocial hehav-ior in children and adolescenis: A devetopnwnml analysis andmodel for inienvntion- Washington. DC: American PsychologicalAssociation.

Rushy, J. C , Estes. A.. & Dishitm. T. (1991). The Inierpersonal Pro-cess Code (fPC). Unpublished m anuscript. Oregon Social Learning

Center, Eugene. OR.Sameroff A. J. (1989}. Comm entary; General systems and the regulationof development. In M. R. Gunnar & E. Thelen (E ds.). Sysietm anddevelopment: The Minnc-soia Symposia on Child Psychology (Vol22 . pp. 219-235). Hillsdale, NJ: Erlbaum.

Shaw. D. S.. Emery. R. E.. Tuer, M.D. (1993). Parental functioning andchild ren's adjustment in fattiilies of divore e: A prospective study.Journul of Almormal Child P.sycliolofiy. 2 1. 119-134.

Sloolmiller. M. S. (1995). Using latent growth curve models to sludydevelopmental processes. In J. M. Gottman (Ed.), The analysis ofchange (pp. I(J5-138). Mahwah. NJ: Erlbaum.

Stoolmiller. M.. Duncan. T. Bank. L.. & Patterson. G- R. (1993). Someproblem s and solutions in the study of ehange: Significant pattern.siii client resistance. Journal ofComulting and Clinical P. vcholog\.61. 92 t~92H.

U.S. Departtiient of Commerce. Bureau of the Census. (1993). 1990

Census of population and lunising: Popiiluiinn and hou.\ing char-(icteristicsfor census tracts an d block numbering areas (1990 CPH-3-144). Washington. DC: U. S. Government Printing Oflice.

Vinokur. A. D.. Price. R. H.. & Caplan. R, D. (1991). From lield ex-periments to program implem entation: Assessing the potential out-comes of an experimental intervention program for unemployedpersons. American .lonnial of Community Ps\cliolot;v. 19. 5 4 i -562.

Vitaro. F.. Brendgen. M.. & Tremblay, R. E. (2001). Preventive inter-vention: Assessing its effects on the trajectories of delinquency andte.sting for mediational processes. AppliedDevelopmenialScietueJ. 201-213.

Webster-Stratton, C . Hollinsworth, T. & KolpacotT, M. (1989). Thelong-term effectiveness and clinical significance of three cost-elTective training programs for families with conduct-problem chil-dren. Journal of Consulting and Clinical P.wchohgv, 57. 550 -.'i.'i3.

Wothke. W. (2(XX)). Longitudinal and multigroup modeling with miss-ing data. In T. D. Little, K. U. Schnabei. & J. Baumert (Eds.),Longitudinal and multilevel data: Praciicul issues, applied ap-proaches, and specific examples (pp. 269-281). Mahwah, NJ:Erlbaum.


14/14

base patterson, 2004 systematic changes in families following prevention trials

Documents